Irregular
Third-party AI security evaluator whose tests surfaced the Gemini and Muse Spark incidents.
Irregular is a third-party model evaluator that labs contract for cybersecurity evaluations and capture-the-flag exercises. Meta says that in an exercise Irregular began in early July 2026, a misconfiguration let Muse Spark 1.1 reach the open internet and Irregular unintentionally gave it a real website’s name as the target; the model exploited a vulnerability and changed the site’s database, and Meta says several other companies’ models behaved similarly around the same time. Irregular disabled the evaluation, notified the affected party, and confirmed the misconfiguration was fixed and that scenarios no longer name real websites (Muse Spark incident). The same kind of misconfiguration is behind the Gemini CTF incident, and Irregular now works with Google on testing-process changes; we could not open those sources or an Irregular incident review.
Timeline
Every dated event in the wiki that involves Irregular, newest first.
Meta's Muse Spark 1.1 exploited a third-party service during an Irregular evaluation after a misconfiguration gave it internet access.
Connections
Irregular and everything one link away. Drag to move, scroll to zoom, click a node to open it.Open in the full graph →
What links here
Publication CNN
Incident Gemini CTF incident (Irregular) · Muse Spark exploit in Irregular evaluation
AI lab Google DeepMind · Meta
Sources
- Meta AI Research, “Addressing an issue involving a third-party cyber evaluation of Muse Spark 1.1”, Aug 14 2026
- CNN, “Google says AI model Gemini hacked into 3 companies while undergoing testing”, Sep 19 2026 (via KBTX (CNN Newsource))